an automated velocity dealiasing scheme for radar data ...22327...ational radar products for the...

An Automated Velocity Dealiasing Scheme for Radar Data Observed fromTyphoons and Hurricanes

GUANGXIN HE

Key Laboratory of Meteorological Disaster, Ministry of Education (KLME), and International Joint Research

Laboratory on Climate and Environment Change (ILCEC), and Collaborative Innovation Center on Forecast and

Evaluation of Meteorological Disasters, and Center of Data Assimilation for Research and Application, Nanjing

University of Information Science and Technology, and Nanjing Joint Center of Atmospheric Research, Nanjing,

and Key Laboratory of Meteorology and Ecological Environment of Hebei Province, Shijiazhuang, China,

and National Center for Atmospheric Research, Boulder, Colorado

JUANZHEN SUN AND ZHUMING YING

National Center for Atmospheric Research, Boulder, Colorado

(Manuscript received 21 September 2017, in final form 15 November 2018)

ABSTRACT

Accurate and automated dealiasing of radar data is important for data interpretation and downstream

applications such as numerical weather prediction (NWP) models. In this paper an improved radial velocity

dealiasing scheme is presented and evaluated using observations from several S-band radars under the severe

weather conditions of typhoons and hurricanes. This scheme, namedAutomatedDealiasing for Typhoon and

Hurricane (ADTH), is a further development of the China New Generation Doppler Weather Radar

(CINRAD) improved dealiasing algorithm (CIDA). The upgraded algorithmADTH includes three modules

designed to select the first radial from which the dealiasing process starts, to conduct a two-way multipass

dealiasing, and to perform an error check for a final local dealiasing. The dealiasing algorithm is applied to two

typhoon hurricane cases and four typhoon cases observed with radars from CINRAD, NEXRAD of the

United States, and the Taiwan radar network for a continuous period of 12 h for each of the selected cases.

The results show that ADTH outperforms CIDA for all of the test cases.

1. Introduction

Doppler radar observations have high spatial and

temporal resolutions and have been used for many me-

teorological applications, ranging from providing oper-

ational radar products for the prediction of adverse

and hazardous weather systems to initializing numerical

weather prediction (NWP) models. Radial velocity ob-

served from Doppler weather radar plays an important

role in radar data assimilation and has considerable po-

tential to improve precipitation predictions (e.g., Sun

2005; Sun and Zhang 2008; Gao et al. 2004; Hu et al. 2006;

Xiao and Sun 2007; Simonin et al. 2014). The assimilation

of these data intoNWPmodels requires accurate and fully

automated quality control of the radar data. One of the

longstanding challenges for the quality control of Doppler

radar radial velocity measurements is the correction of

aliased velocities.

The maximum observed velocity, or Nyquist velocity

Vmax, from the Doppler radar data is related to the pulse

repetition frequency (PRF) of the radar with a trans-

mitted wavelength l. It is defined as

Vmax

5(PRF)l

4. (1)

The interval [2Vmax, Vmax] refers to the unambiguousvelocity interval. As shown in Eq. (1), Vmax can be in-

creased by increasing the PRF. Nevertheless, since the

maximum unambiguous range Rmax is inversely pro-

portional to the PRF, a larger PRF will reduce the Rmax.

The trade-off between the Vmax and Rmax leads to the

limited values of the Nyquist velocity. Any real radial

velocity not in the interval of [2Vmax, Vmax] will appearwithin 6Vmax and thus will be aliased. The relationshipCorresponding author: Dr. G. He, [email protected]

JANUARY 2019 HE ET AL . 139

DOI: 10.1175/JTECH-D-17-0165.1

� 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS CopyrightPolicy (www.ametsoc.org/PUBSReuseLicenses).

mailto:[email protected]://www.ametsoc.org/PUBSReuseLicenseshttp://www.ametsoc.org/PUBSReuseLicenseshttp://www.ametsoc.org/PUBSReuseLicenses

between the true radial velocity VT and the observed

velocity VO is as follows:

VT5V

O1 2n3V

max, (2)

where n is an integer. An effective radial velocity deal-

iasing algorithm can recover VT from the raw radar

observation VO by choosing the proper n.

Since the 1970s, many dealiasing algorithms have been

developed. These algorithms are mainly based on two gen-

eral techniques, namely, the continuity check method and

the referencewindcheckmethod.Although somesuccesshas

been demonstrated, accurate velocity dealiasing remains a

challenge. Inparticular,when the radial velocity is assimilated

into an operational NWP model, it is necessary to have a

dealiasing scheme that is fully automated and highly reli-

able. A small percentage of failed or erroneously dealiased

data can cause analysis errors in themodel’s initial conditions

and hence contaminate the subsequent forecasts.

The reference wind check method uses a reference

wind field to dealias the radial velocity at each grid point

individually (Gao et al. 2004; Lim and Sun 2010; Xu et al.

2011). As such, it has the advantage of being unaffected

by any erroneously dealiased data of its neighboring

points, avoiding the spatial spreading of such an error,

which can occur in algorithms based on the continuity

check method. However, its disadvantage is that the

method can fail for grid points where the reference

winds have large errors. The first continuity check al-

gorithm, introduced byRay andZiegler (1977), was one-

dimensional, in which only radial gate information was

used to correct the aliased velocities along each radial.

Following their work, several two-dimensional algo-

rithms (Merritt 1984; Bergen and Albers 1988; Eilts and

Smith 1990; Jing andWiener 1993; Gong et al. 2003; Gao

et al. 2004; Zhang and Wang 2006; He et al. 2012b) that

dealiased in both the azimuthal and radial directions

were developed. Some algorithms were combined with a

reference wind check method in which additional in-

formation from a radiosonde or velocity–azimuth dis-

play (VAD) wind profile (Tabary et al. 2001) was used

for the dealiasing of isolated areas (Merritt 1984; Eilts

and Smith 1990; Gong et al. 2003; Gao et al. 2004;

Xu et al. 2011). Based on a two-dimensional dealiasing

algorithm, James and Houze (2001) developed a four-

dimensional algorithm by adding information from dif-

ferent elevation angles and volume scans. Most of the

algorithms are designed for a Nyquist velocity between

20 and 36m s21 but tend to fail for some typhoon and

hurricane cases when multiple-aliased velocity obser-

vations appear. A series of algorithms (Wang et al. 2012;

Xu et al. 2014; Jiang and Xu 2016) were developed for

aliased radar radial velocities obtained from a hurricane

and a typhoon. The algorithms by Xu et al. (2014) and

Jiang and Xu (2016) are built on alias-robust velocity–

azimuth display analysis (Xu et al. 2010).

He et al. (2012a, hereafter HE2012a) proposed an up-

graded two-dimensional continuity check scheme for

dealiasing the S-band China New Generation Doppler

Weather Radar (CINRAD) observations, named the

CINRAD improved dealiasing algorithm (CIDA). A

novel feature of CIDA was the use of a new method to

select the initial reference radial along a ‘‘zero line’’

where the radial velocities are very small and less likely to

be aliased. In addition, a multipass procedure in CIDA

enabled the improved dealiasing of aliased velocity ech-

oes adjacent to missing data, which also eliminated the

need for VAD wind profiles or radiosonde. The appli-

cations of the algorithm to multiple cases of typhoons,

squall lines, and heavy rains observed by CINRAD

showed superior performance to those of two standard

NEXRAD (U.S. radar network) algorithms (HE2012a).

Although HE2012a has made significant contribu-

tions toward our goal to develop a fully automated,

accurate radar velocity dealiasing scheme that is appli-

cable for different weather system observation, the

question remains whether the scheme is generally ap-

plicable for radar measurements from all the typhoon

and hurricane cases. Also, HE2012a tested CIDA on

only one selected volume for each of the selected

weather cases. Our experience with CIDA showed that

the algorithm could fail in some typhoon and hurricane

situations when, for example, multiple aliases occurred,

the typhoon or hurricane eye was very close to radar

station, or aliased velocities were surrounded by missing

data. Therefore, improvements are required for the

algorithm to be capable of and robust enough for day-

to-day real-time observations. In the current study, we

improved the dealiasing algorithm of HE2012a and it is

referred to as Automated Dealiasing for Typhoon and

Hurricane (ADTH). We implemented improvements in

three areas with the goal of developing a fully auto-

mated, accurate radar velocity dealiasing scheme for

real-time typhoon and hurricane observations. ADTH

was applied to radar data from several different typhoon

and hurricane cases to examine its robustness. We ran

ADTH continuously for 12 h for each of the selected

cases to evaluate the scheme’s robustness with large

samples of observations and to obtain statistical verifi-

cation. In these cases, more than 75% of the raw ob-

served data have aliased velocities. To quantify the

improvement of ADTHover CIDA, all of the same data

were run through CIDA.

TheADTHdescribed in this paper includes three new

developments. First, the refinement of the first radial

selection method using winds estimated from the radial

140 JOURNAL OF ATMOSPHER IC AND OCEAN IC TECHNOLOGY VOLUME 36

velocity as an additional constraint; second, the upgra-

ded multipass dealiasing algorithm added passes along

the azimuthal direction; third, an improved least squares

error check method to perform localized dealiasing.

The rest of the paper is organized as follows. Section 2

describes the details of themodules in theADTHalgorithm.

Section3provides thedealiasing results of four typhooncases

and two hurricane cases and the results of their statistical

analyses. Finally, section 4 concludes with a brief summary.

2. Description of the ADTH algorithm

The improved dealiasing algorithm ADTH consists of

threemodules: 1) Selection of the first radial. This criterion

is based on the gradient velocity–azimuth display (GVAD)

wind to reduce the failure rate in typhoon/hurricane

cases, especially when multiple-aliased velocities appear. 2)

Dealiasing in radial and azimuthal directions. This

includes a two-way dealiasing procedure in both radial and

azimuthal directions to improve the dealiasing reliability,

especially when the typhoon/hurricane eye is very close

to a radar station. 3) The third is local error checks. This

includes a local error check and a linear square error check

in the radial direction and a quadratic least squares error

check in the azimuthal direction. These three modules are

described in detail below. Attention will be paid to our

improvements over the original scheme in HE2012a in the

following description. Among the three modules, modules

1 and 3 are most sophisticated with new methodologies;

thus, three figures (Figs. 1 and 4, and 5) are provided for

these modules to facilitate the description.

a. Module 1: Selection of the first radial

A CRITERION BASED ON THE GVAD WIND TOREDUCE THE FAILURE RATE IN TYPHOON/HURRICANE CASES, ESPECIALLY WHENMULTIPLE-ALIASED VELOCITIES APPEAR

With the limitation of hardware performance of the

radar station, there will be some range folding gates

appearing in low-elevation angles of the radial velocity

field. All of the range folding from different radar ob-

servations will be abandoned and set up with missing

data before dealiasing. This step will not change the

value of valid gates and will not affect the performance

of our dealiasing algorithm.

Correctly selecting the first radial such that the radial

velocities do not contain aliased data is critical for deal-

iasing algorithms to work successfully, as the subsequent

dealiasing depends on the previously corrected radial

velocity. Because the radial velocities along the radial

where the atmospheric wind is nearly perpendicular to

the radar beam are very small (or near zero) and are

therefore less likely to be aliased, CIDA finds the zero

line (illustrated by the yellow line in Figs. 1a and 1b) by

computing the mean radial velocity averaged along each

radial (blue curves in Figs. 1c and 1d). Also plotted in

Figs. 1c and 1d are the total numbers of the valid gates

used in the calculations of the means (red curves). The

first radial should have a local minimum average of mean

radial velocities and the largest number of valid gates

along each radial. Because the near-zero radial velocity

typically appears at two radials (the two local minima

marked by L1 with 382 valid gates and L2 with 547 valid

gates in Fig. 1c), the radial with more valid range gates

(L2) is selected as the first radial, which is at ;2608, asshown by the yellow line in Fig. 1a.

The above scheme in CIDA, originally described in

HE2012a, works well in determining the proper first ra-

dial in most situations but tends to fail when multiple-

aliased velocities appear in some typhoon or hurricane

cases. As an example, Fig. 1b shows a failure case from an

S-band Taiwan radar system with an elevation angle of

19.58 at 2208 UTC 18 September 2010 from the TyphoonFanapi observations from the radar located in Hualian,

Taiwan. More than 80% of the radial velocities are

aliased at this elevation. The yellow line (corresponding

to L1 in Fig. 1d) is not the best first reference radial found

using the original method in CIDA, which was due to the

second aliasing, which resulted in small values of aliased

velocities between the 308 and 458 azimuth angles.To eliminate the ambiguities caused by the contami-

nations of the radial velocity data, we added a criterion in

ADTH based on the prevailing environmental wind di-

rection. The wind is estimated using the GVAD method

developed by Gao et al. (2004). The classic VAD tech-

nique (Browning andWexler 1968) is able to estimate the

areal mean vertical profile of the horizontal wind above

a Doppler radar. However, the classic VAD technique

encounters difficulty when the radial velocities are aliased

or have large areas of missing data. The GVAD tech-

nique uses velocity gradient information to retrieve the

wind speed and direction from radial velocity observa-

tions without the need for prior dealiaising and other

measurements. The GVAD technique can calculate the

wind components u0 and y0 at each level z0 for a constant

scan elevation angle through the minimization of the

following cost function:

J(u0, y

0)5 �

N

n51

1

s2n

��›V

o

›u

�n

2 cosun

3 [(cosun)u

02 (sinu

n)y

0]

�2, (3)

where u is the azimuth angle, u is the elevation angle,sn 5 0:8 represents the quality of the fit by calculating


the dispersion of the data around the straight line, and n is

the index of the data in the azimuthal direction (Tabary

et al. 2001). The calculation of the azimuthal gradients of

radial velocity is similar to Eq. (11) in Gao et al. (2004)

and is done by taking the difference of the velocity at

adjacent points in the azimuthal direction,

�›V

o

›u

�n

5

�Vn11o 2V

no

un11 2un

�. (4)

Now we illustrate how the GVAD wind is used as an

additional criterion in determining the first radial using

the example shown in Fig. 1b. In the first step, four ra-

dials with local minima, marked by L1, L2, L3, and L4,

respectively, in Fig. 1d, were determined. Among the

four local minima, L1 was chosen as a candidate first

radial because it had the smallest average radial velocity

and the largest number of valid data points. Using the

least squares fit method in Eq. (3), the average wind

direction over all the vertical levels for the elevation

angle in Fig. 1b is found to be 202.78 (northward wind is08 here). Hence, the zero lines should be near 112.78 or292.78 (perpendicular directions to that of the wind),which correspond to the two local minima, L2 and L4, in

Fig. 1d.We choose L2 as the possible first radial because

it has a smaller mean radial velocity than L4 (see

Fig. 1d). For the final step to determine whether L1 or

L2 is the real zero line, the following criterion is used.

If the difference between L1 and L2 is less than 208(empirically determination), suggesting that L1 is near

the zero line, then it is safe to choose the radial L1 as the

first radial; otherwise, the radial L2 is selected.

The GVAD wind criterion is necessary for correctly

dealiasing typhoon and hurricane cases. For more de-

tails on the basic procedure of the first radial selection,

the reader is referred to HE2012a.

FIG. 1. Raw radar radial velocity (a) at 1.58 elevation angle from the CINRADS-band radar located inWenzhou,China, at 1045 UTC 29 Jul 2008 and (b) at 19.58 elevation angle from the S-band radar located at Hualian at2208 UTC 18 Sep 2010. MD is the abbreviation of missing data observation. The yellow lines are the first reference

radial using the method in HE2012a, and the red line is the reference radial calculated by the GVAD method.

(c),(d) The radial-averaged mean radial velocity with respect to azimuth angle (blue curve with north as 08, left y axis,and m s21) and the total number of the valid gates used for the computation of the mean (red curve, right y axis) for

(a) and (b), respectively.


b. Module 2: Multipass dealiasing

A TWO-WAY DEALIASING PROCEDURE IN BOTHRADIAL AND AZIMUTHAL DIRECTIONS TO

IMPROVE THE DEALIASING RELIABILITY,ESPECIALLY WHEN TYPHOON/HURRICANE EYE ISVERY CLOSE TO RADAR STATION

Dealiasing starts with the radials next to both sides of

the first reference radial chosen by module 1 in the two

passes: one in the clockwise direction and the other in the

counterclockwise direction. Each pass goes through 1808.In each of the half circles, the module dealiases in the

radial and azimuthal directions in a two-way fashion (see

the illustration in Fig. 2). The procedure in ADTH is

similar to that in HE2012a but with the additional steps

along the azimuthal direction (Figs. 2c and 2d). For each

gate, the dealiasing is performed using a continuity check

that compares its radial velocity with a reference radial

velocity, which is selected from the dealiased velocities

in the preceding gates. For the details of how to select

the reference radial velocity, the reader is referred to

HE2012a. This two-way and two-direction (radial and

azimuthal) dealiasing procedure does not have the de-

pendence of sounding or VAD wind profiles and is ef-

fective for dealing with near-range aliasing and aliasing

next to missing gates.

c. Module 3: Local error check and dealiasing

A LOCAL ERROR CHECK AND A LINEAR SQUAREERROR CHECK IN RADIAL DIRECTION AND A

QUADRATIC LEAST SQUARES ERROR CHECK IN

AZIMUTHAL DIRECTION TO CLEAN UP THE

REMAINING DEALIASING ERRORS

A local error check is performed as a final step inADTH

to clean up any local errors that may still exist after the

abovementioned two modules. At each gate, an average

velocity of V over all the valid points k within a specified

geometrical window is calculated as the reference wind for

the following equation to compute a suitable n in Eq. (2):

n5NINT

V2V

o

2Vmax

!, (5)

where NINT represents rounding off to the nearest

integer. The size of the geometrical window is defined

by the number of gates k, which is empirically chosen and

varies with the range distance from a radar station. The

size of the window is set to 93 9 and 153 15 for distancesbetween the radar station and the observation gate less

than and greater than 100km, respectively. The local

error checks and dealiasing are not performed when the

number of valid points in the specifiedwindow is less than

70% (empirically determined) of the total gates to avoid

uncertainty.

The above local dealiasing scheme works successfully for

most weather situations, but it may fail when the eye of a

hurricane or typhoon is close to the radar station. Therefore,

the following two steps are added in ADTH to ensure the

robustness of the algorithm.First, a linear least squares error

check in the radial direction is carried out. The velocity is

assumed to satisfy a linear relationship along each radial:

VT 5 am 3Xn 1 bm, where VT is the true radial velocityand Xn is the distance between a range gate and the radar

station. The constants am and bm are obtained by solving a

linear least squares equation using the observed radial ve-

locities on that radial. All the velocity values that do not fit

this constraint well are then dealiased using Eq. (2). The

second step is to perform a quadratic least squares error

check in the azimuthal direction. Since the measured radial

velocity is the wind velocity projected onto the radial di-

rection, in each half circle, starting from thefirst radial l1 and

extending to the radial l2 (1808 away from l1), all the ve-locities with the same range from the radar in the azimuthal

direction approximately satisfy a quadratic relationship

(the sine function between 08 and6p). Thus, a quadraticleast squares fit is used to identify and correct any re-

maining aliased data.

3. Results of ADTH applications

The dealiasing algorithm ADTH described above was

tested on four typhoon cases and twohurricane cases. These

six cases were observed by S-band radar from different ra-

dar stations. The radial velocity observations covered a ra-

dius of 230km for the S-band radar systems. Depending on

thePRFsof the different radars, theNyquist velocity ranged

from 21.15 to 31.61ms21. The ADTH’s continuous deal-

iasing covered a period close to 12h for each case. And

more than half of the raw volumes contained aliased ve-

locities in each selected case. The total number of volumes

was 789 and the number of constant-elevation scans was

4847. Table 1 presents the total volumes, the total constant-

elevation scans, and the total valid gates of the six selected

periods. Table 1 also shows the number of volumes con-

taining data with aliased velocities and the number of gates

with aliased velocities in raw data in each period, of which

75% of volumes have aliased velocities and 44.9% of the

gates contain aliased velocities between the six selected

cases. The dealiasing results for these cases are shown

below and are followed by a statistical evaluation.

a. Result of typhoon case

Typhoon Fanapi formed east of the Philippines, moved

westward, and struck Taiwan on 19 September 2010.


It was notable because of the severe property damage and

five lost lives that occurred in Taiwan alone as a result of

heavy rainfall, landslides, and floods. With a maximum

sustained wind speed of 54ms21, some Taiwan radar

systems experienced double aliasing, which made the

dealiasing more challenging. Another challenge for ty-

phoon dealiasing can occur when a typhoon eye is close

to a radar station because the radial velocity field exhibits

asymmetry at the two sides of the zero-velocity line.

The S-band radar located in Hualian near the eastern

coast of Taiwan has a relatively low Nyquist velocity of

21.15ms21 andwas close to the typhoon eye at 2208UTC

18 September 2010, right before Fanapi made landfall.

Therefore, it provides good datasets to test ADTH’s ca-

pability in extreme circumstances. Figures 3a, 3c, and 3e

show the raw radial velocities observed by the Hualien

radar at the elevation angles of 1.58, 4.38, and 19.58, re-spectively. Because of the blocking by themountains near

the coastline of Taiwan, the Hualien radar operates by

only partial scanning over the elevation angles between

0.58 and 9.98 and fully scanning the two high-elevationangles above. The partial scanning coverage of the wind

fields at low-elevation angles can cause particular diffi-

culties for any dealiasing algorithms.As seen in Fig. 3, the

eye of Fanapi was only 50km east of the radar at this

time. The velocity observations to the northwest and

southwest of the eye of the typhoon show considerable

twice-aliased data. ADTH correctly detected the first

radial near the zero line and corrected the twice-aliased

data. Figure 3b is the final dealiased result at a 1.58 el-evation angle after all of the dealiasing modules were

run (the color scales are different between before and

after dealiasing of this case). At the 4.38 elevation angle,more than 90% of the valid radial velocities were

aliased and there exists a discontinuous echo to the

north of the typhoon eye as shown in Fig. 3c. For this

elevation angle, the local dealiasing method based on

the least squares error check described in the last

module played an important role. As shown in Fig. 3d,

ADTH successfully corrected the severely aliased ve-

locity data. For the elevation angle of 19.58, the GVADtechnique in module 2 proved useful in correctly finding

the first reference radial near 2928 in Fig. 3e, as detailedin the description of Fig. 1 in section 2. The correct

first reference radial ensured the rest of the modules

were executed successfully, resulting in the successful

dealiasing of this angle as shown in Fig. 3f.

b. Result of hurricane case

Hurricane Isaac made landfall in southeastern Louisi-

ana on 21 August 2012. It caused extensive heavy rainfall

and inland flooding over southern Mississippi and

southeastern Louisiana. Isaac was estimated to be di-

rectly responsible for five deaths in the United States.

FIG. 2. An illustration of dealiasing along the radial direction (a) from inner to outer range gate and (b) from

outer to inner range gate, and along the azimuthal direction proceeding (c) from inner to outer range gate and

(d) from outer to inner range gate.


The S-band KLIX radar, located in New Orleans,

Louisiana, at 2043UTC 28August 2012, right before Isaac

made landfall near Louisiana, provided good datasets

to evaluate ADTH’s performance when the gates with

aliased velocities are surrounded by gates with missing

data, which causes difficulty for any dealiasing algorithm.

Figure 4a shows the raw radial velocity at the elevation

angle of 0.58. In spite of the existence of the hurricane eyeand the discontinuous velocity data near the aliased area to

the south and southeast of the radar, the new algorithm

performs very well by correctly dealiasing two areas in

Fig. 4b. Figures 4c and 4d show the raw radial velocity and

dealiased radial velocity, respectively, at the 4.38 elevationangle at the same observation time. Again, the dealiased

velocity field is clean and without noticeable errors.

The above results from some selected scans demon-

strated that the newADTH algorithm has the capability

to address challenging aliasing issues in hurricane situ-

ations. The overall statistics for both hurricane cases

over a period of 12 h are computed, and the results will

be discussed in the next subsection.

c. Comparison with the original CIDA via statisticalevaluations

The skill of ADTH over all the tested radar volumes

from the four typhoon cases and the two hurricane cases

was evaluated statistically and compared with that of

CIDA, as described inHE2012a. The evaluationwas based

on the percentage of gates with correct velocity data after

running two different algorithms. The period of continuous

dealiased data is close to 12h for each case. There are 75%

raw observed volumes containing gates with aliased ve-

locities in all the selected periods, which means most of the

volumes contain velocity data that need to be dealiased.All

of the constant-elevation scans were examined, and the

fields with aliased velocities were manually edited to pro-

duce ‘‘truths’’ for evaluation. The truths radial velocities

were manually edited by the National Center for Atmo-

spheric Research (NCAR) SOLO II package (HE2012a).

Since the performance statistics depend on the accuracy of

the manual dealiasing process, we carefully checked the

manual editing results to detect and correct any possible

mistakes. Given our expertise and experience in the area,

weare confident that themanual editing results are reliable.

Table 2 shows the statistical dealiasing results in-

dicating the percentage of all correct gates after deal-

iasing out of all gates with valid data for the six cases

using ADTH and CIDA. It is shown that ADTH out-

performs CIDA for each case. The Taiwan Hualian ra-

dar for the Typhoon Fanapi case and the China Haikou

radar for the Typhoon Rammasun case had the largest

skill increase, which was mainly attributable to the im-

proved detection of the first radial and special local error

TABLE1.Summary

ofthesixstudycases;N:NEXRAD,TW

:TaiW

ai,C:CIN

RAD,H:hurricane,T:typhoon.KAKQ

islocatedin

Wakefield,VA.

Radar

site

KLIX

(N)

KAKQ

(N)

Hualien(T

W)

Haikou(C

)W

enzhou(C

)Xiamen

(C)

Total

Weathersystem

Isaac(H

)Irene(H

)Fanapi(T

)Ram

masun(T

)Fungwong(T

)Morakot(T

)

Period

1100–2300UTC

28Aug2012

1400UTC27Aug–0200

UTC

28Aug2011

1000–2200

UTC

18Sep2011

2100UTC17

Jul–1100UTC

18Jul2014

0100–1400

UTC

29Jul2008

0300–1500

UTC

8Aug2008

Numberofvolumes

156

147

89

144

131

122

789

Volumes

withaliasedvelocities

122(78.2%

)85(57.8%

)78(87.6%

)118(81.9%

)95(72.5%

)94(77.0%

)592(75.0%

)

Numberofelevationscans

1376

1320

712

1296

1179

1098

7981

Gateswithvalidvelocities

643946

603007

353872

518408

462919

420971

3003123

Gateswithdealiasedvelocities

212507(33.0%

)229855(38.1%

)236038(66.7%

)269381(51.9%

)198325(42.8%

)202934(48.2%

)1349040(44.9%

)


FIG. 3. Radar radial velocity from the Hualien radar of the Taiwan radar network at 2208 UTC 18 Sep 2010.

(a) Raw velocity at 1.58 elevation angle, (b) dealiased result at 1.58 elevation angle, (c) raw velocity at 4.38 elevationangle, (d) dealiased result at 4.38 elevation angle, (e) raw velocity at 19.58 elevation angle, and (f) dealiased result at19.58 elevation angle. The Nyquist velocity for theHualien radar is 21.15m s21. The x axis is the distance away fromthe radar in thewest–east direction, while the y axis is the distance away from the radar in the south–north direction.

The spatial scales of the pairs of images at different elevations are different. The yellow lines are the first reference

radials.


checks. And more than half of the aliased data are twice

aliased, which the CIDA cannot deal with well in these two

cases. Comparing the statistical results of the six cases, the

new ADTH algorithm can dealias more than 98% of the

radial velocity correctly. It has a reliable performance for

the typhoon and hurricane systems analyzed here.

4. Summary and conclusions

In this paper we presented and evaluated the velocity

dealiasing algorithmADTH, which was developed as an

improvement to CIDA (operational on the Chinese

radar network) with the aim of dealiasing typhoon and

hurricane systems. The performance of the ADTH al-

gorithm was examined for four typhoon cases and two

hurricane cases observed by S-band radars stations, to-

taling 789 volumes or 7981 constant-elevation velocity

fields. The selected cases had raw velocities that were

severely aliased for a large percentage of the constant-

elevation scans, and some included challenging deal-

iasing situations with multiple-aliased radial velocities

as a result of strong wind speeds.

FIG. 4. Radar radial velocity from KLIX of NEXRAD at 2043 UTC 28 Aug 2012. (a) Raw velocity at 0.58elevation angle, (b) dealiased result at 0.58 elevation angle, (c) raw velocity at 4.38 elevation angle, and (d) dealiasedresult at 4.38 elevation angle. The spatial scales of the pairs of images at different elevations are different. Theyellow lines are the first reference radials.

TABLE 2. List of all six cases and the associated dealiasing results (the percentage of all correct gates after dealiasing out of all valid gates)

using the ADTH algorithm and the original CIDA algorithm; H: hurricane, T: typhoon.

Radar site KLIX (H) KAKQ (H) Hualien (T) Haikou (T) Wenzhou (T) Xiamen (T) Total

ADTH 99.04% 99.13% 98.32% 98.91% 99.62% 99.54% 99.02%

CIDA 97.36% 96.68% 88.79% 80.33% 93.11% 92.75% 92.90%


The new ADTH comprises three modules that were

designed to determine the starting reference radial, to

identify andmodify the aliased velocity gates with a two-

way multipass scheme and to rectify the incorrect results

with a final local error check. The selection of the first

radial near the zero-velocity line ensures that the deal-

iasing algorithm starts from velocity data that are less

likely aliased. The GVAD technique, using the velocity

gradient information in module 1, was shown to help

improve the accuracies of the first radial detections in

the typhoon and hurricane cases where the multiple-

aliased situations occur rather frequently. The two-way

multipass dealiasing method implemented in module 2

improved the accuracy of the dealiasing, especially when

the aliasing occurs near large areas of missing data. In

addition, we showed that the additional linear least

squares error check in the radial direction and quadratic

least squares error checks in the azimuthal direction

were able to eliminate the remaining dealiasing errors in

the typhoon and hurricane systems.

The statistical evaluation of ADTH, using all correct

gates after dealiasing out of all valid gates, demonstrated

that the upgraded ADTH outperformed the preceding

algorithm of CIDA for all six cases, with largest im-

provements for the Taiwan Typhoon Fanapi case and for

the China Typhoon Rammasun case, which contain lots

of twice-aliased observed velocity gates. The improve-

ment resulted from more accurate detections of the first

radial and local error checks. The success rate of ADTH

for the typhoon and hurricane systems was above 98%.

ADTH is developed for dealiasing typhoon and hurri-

cane systems and is applied only when these types of

storms are already present. The fact is that practical

and operational application of the ADTH requires human

intervention or a special algorithm that will help to de-

termine the typhoon and hurricane storms. Lee andMarks

(2000) and Lee et al. (2014) have developed a GBVTD-

simplex algorithm that can reduce the uncertainties in es-

timating the TC center position and improves the quality

of the GBVTD-retrieved TC circulation. The GBVTD-

simplex algorithm is computationally efficient for real-time

applications. Such algorithms like the GBVTD-simplex

algorithm can be used to identify typhoon and hurricane

storms and to decide whether to execute ADTH for

dealiasing. ADTH is being implemented for the Varia-

tional Doppler Data Analysis System (VDRAS) and the

Weather Research and Forecasting Data Assimilation

(WRFDA) system as one of the radar quality control

software systems. Both systems have operational capabil-

ities, so ADTH can be tested operationally in the future.

Acknowledgments. This work was supported by the

ResearchProgram forNanjing JointCenter ofAtmospheric

Research (NJCAR2016MS01), the China Special Fund for

Meteorological Research in the Public Interest (Grant

GYHY 201506004), the Research Program for Key Labo-

ratory of Meteorology and Ecological Environment of

Hebei Province (Z201603Z), Key Laboratory of South

China Sea Meteorological Disaster Prevention and Miti-

gation of Hainan Province (Grant SCSF201804), and the

PriorityAcademicProgramDevelopment of JiangsuHigher

Education Institutions (PAPD).

REFERENCES

Bergen,W. R., and S. C. Albers, 1988: Two- and three-dimensional

dealiasing of Doppler radar velocities. J. Atmos. Oceanic

Technol., 5, 305–319, https://doi.org/10.1175/1520-0426(1988)

005,0305:TATDDA.2.0.CO;2.Browning, K. A., and R. Wexler, 1968: The determination of

kinematic properties of a wind field using Doppler radar. J. Appl.

Meteor., 7, 105–113, https://doi.org/10.1175/1520-0450(1968)

007,0105:TDOKPO.2.0.CO;2.Eilts, M. D., and S. D. Smith, 1990: Efficient dealiasing of Doppler ve-

locities using local environment constraints. J. Atmos. Oceanic


007,0118:EDODVU.2.0.CO;2.Gao, J., K. K. Droegemeier, J. Gong, and Q. Xu, 2004: A method for

retrieving mean horizontal wind profiles from single-Doppler

radar observations contaminated by aliasing. Mon. Wea. Rev.,

132, 1399–1409, https://doi.org/10.1175/1520-0493-132.1.1399.Gong, J., L. Wang, and Q. Xu, 2003: A three-step dealiasing method

for Doppler velocity data quality control. J. Atmos. Oceanic


020,1738:ATDMFD.2.0.CO;2.He, G., G. Li, X. Zou, and P. S. Ray, 2012a: Applications of a ve-

locity dealiasing scheme to data from China New Generation

Weather Radar System (CINRAD). Wea. Forecasting, 27,

218–230, https://doi.org/10.1175/WAF-D-11-00054.1.

——,——,——, and——, 2012b: A velocity dealiasing scheme for

synthetic C-band data fromChina’s NewGenerationWeather

Radar System (CINRAD). J. Atmos. Oceanic Technol., 29,

1263–1274, https://doi.org/10.1175/JTECH-D-12-00010.1.

Hu, M., M. Xue, and K. Brewster, 2006: 3DVAR and cloud analysis

with WSR-88D level-II data for the prediction of the Fort Worth,

Texas, tornadic thunderstorms. Part I: Cloud analysis and its impact.

Mon. Wea. Rev., 134, 675–698, https://doi.org/10.1175/MWR3092.1.

James, C. N., and R. A. Houze Jr., 2001: A real-time four-

dimensional Doppler dealiasing scheme. J. Atmos. Oceanic

Technol., 18, 1674–1683, https://doi.org/10.1175/1520-0426

(2001)018,1674:ARTFDD.2.0.CO;2.Jiang, Y., and Q. Xu, 2016: Adaptive dealiasing for Doppler velocities

scanned from hurricanes and typhoons. J. Atmos. Oceanic Tech-

nol., 33, 1931–1947, https://doi.org/10.1175/JTECH-D-15-0146.1.

Jing, Z., andG.Wiener, 1993: Two-dimensional dealiasing of Doppler

velocities. J.Atmos.Oceanic Technol., 10, 798–808, https://doi.org/

10.1175/1520-0426(1993)010,0798:TDDODV.2.0.CO;2.Lee, W., and F. D. Marks, 2000: Tropical cyclone kinematic

structure retrieved from single-Doppler radar observations.

Part II: The GBVTD-simplex center finding algorithm. Mon.

Wea. Rev., 128, 1925–1936, https://doi.org/10.1175/1520-0493

(2000)128,1925:TCKSRF.2.0.CO;2.——, X. Tang, and B. J. Jou, 2014: Distance velocity–azimuth dis-

play (DVAD)—New interpretation and analysis of Doppler


https://doi.org/10.1175/1520-0426(1988)0052.0.CO;2https://doi.org/10.1175/1520-0426(1988)0052.0.CO;2https://doi.org/10.1175/1520-0450(1968)0072.0.CO;2https://doi.org/10.1175/1520-0450(1968)0072.0.CO;2https://doi.org/10.1175/1520-0426(1990)0072.0.CO;2https://doi.org/10.1175/1520-0426(1990)0072.0.CO;2https://doi.org/10.1175/1520-0493-132.1.1399https://doi.org/10.1175/1520-0426(2003)0202.0.CO;2https://doi.org/10.1175/1520-0426(2003)0202.0.CO;2https://doi.org/10.1175/WAF-D-11-00054.1https://doi.org/10.1175/JTECH-D-12-00010.1https://doi.org/10.1175/MWR3092.1https://doi.org/10.1175/1520-0426(2001)0182.0.CO;2https://doi.org/10.1175/1520-0426(2001)0182.0.CO;2https://doi.org/10.1175/JTECH-D-15-0146.1https://doi.org/10.1175/1520-0426(1993)0102.0.CO;2https://doi.org/10.1175/1520-0426(1993)0102.0.CO;2https://doi.org/10.1175/1520-0493(2000)1282.0.CO;2https://doi.org/10.1175/1520-0493(2000)1282.0.CO;2

velocity. Mon. Wea. Rev., 142, 573–589, https://doi.org/10.1175/

MWR-D-13-00196.1.

Lim, E., and J. Sun, 2010:A velocity dealiasing technique using rapidly

updated analysis from a four-dimensional variational Doppler

radar data assimilation system. J. Atmos. Oceanic Technol., 27,

1140–1152, https://doi.org/10.1175/2010JTECHA1300.1.

Merritt, M. W., 1984: Automatic velocity dealiasing for real-time

applications. Proc. 22nd Conf. on Radar Meteorology, Zurich,

Switzerland, Amer. Meteor. Soc., 528–533.

Ray, P. S., and C. Ziegler, 1977: De-aliasing first-moment Doppler

estimates. J. Appl. Meteor., 16, 563–564, https://doi.org/

10.1175/1520-0450(1977)016,0563:DAFMDE.2.0.CO;2.Simonin, D., S. P. Ballard, and Z. Li, 2014: Doppler radar radial

wind assimilation using an hourly cycling 3D-Var with a

1.5 km resolution version of the Met Office Unified Model for

nowcasting. Quart. J. Roy. Meteor. Soc., 140, 2298–2314,

https://doi.org/10.1002/qj.2298.

Sun, J., 2005: Initialization and numerical forecasting of a supercell

storm observed during STEPS.Mon. Wea. Rev., 133, 793–813,https://doi.org/10.1175/MWR2887.1.

——, and Y. Zhang, 2008: Analysis and prediction of a squall line

observed during IHOP using multiple WSR-88D observa-

tions.Mon.Wea. Rev., 136, 2364–2388, https://doi.org/10.1175/2007MWR2205.1.

Tabary, P.,G. Scialom, andU.Germann, 2001:Real-time retrieval of

the wind from aliased velocities measured by Doppler radars.

J. Atmos. Oceanic Technol., 18, 875–882, https://doi.org/10.1175/

1520-0426(2001)018,0875:RTROTW.2.0.CO;2.Wang, M., K. Zhao, W.-C. Lee, B. Jong-Dao Jou, and M. Xue,

2012: The gradient velocity track display (GrVTD) tech-

nique for retrieving tropical cyclone primary circulation

from aliased velocities measured by single-Doppler radar.

J. Atmos. Oceanic Technol., 29, 1026–1041, https://doi.org/

10.1175/JTECH-D-11-00219.1.

Xiao, Q., and J. Sun, 2007: Multiple-radar data assimilation and

short-range quantitative precipitation forecasting of a squall

line observed during IHOP_2002. Mon. Wea. Rev., 135,

3381–3404, https://doi.org/10.1175/MWR3471.1.

Xu, Q., K. Nai, and L. Wei, 2010: Fitting VAD winds to aliased

radial-velocity observations: A global minimization problem

in the presence of multiple local minima. Quart. J. Roy. Me-

teor. Soc., 136, 451–261, https://doi.org/10.1002/qj.589.

——, ——, ——, P. Zhang, S. Liu, and D. Parrish, 2011: A VAD-

based dealiasing method for radar velocity data quality con-

trol. J. Atmos. Oceanic Technol., 28, 50–62, https://doi.org/10.1175/2010JTECHA1444.1.

——,Y. Jiang, and L. Liu, 2014: Fitting parametric vortices to aliased

Doppler velocities scanned from hurricanes.Mon. Wea. Rev.,

142, 94–106, https://doi.org/10.1175/MWR-D-12-00362.1.Zhang, J., and S.Wang, 2006: An automated 2DmultipassDoppler

radar velocity dealiasing scheme. J. Atmos. Oceanic Technol.,

23, 1239–1248, https://doi.org/10.1175/JTECH1910.1.

https://doi.org/10.1175/MWR-D-13-00196.1https://doi.org/10.1175/MWR-D-13-00196.1https://doi.org/10.1175/2010JTECHA1300.1https://doi.org/10.1175/1520-0450(1977)0162.0.CO;2https://doi.org/10.1175/1520-0450(1977)0162.0.CO;2https://doi.org/10.1002/qj.2298https://doi.org/10.1175/MWR2887.1https://doi.org/10.1175/2007MWR2205.1https://doi.org/10.1175/2007MWR2205.1https://doi.org/10.1175/1520-0426(2001)0182.0.CO;2https://doi.org/10.1175/1520-0426(2001)0182.0.CO;2https://doi.org/10.1175/JTECH-D-11-00219.1https://doi.org/10.1175/JTECH-D-11-00219.1https://doi.org/10.1175/MWR3471.1https://doi.org/10.1002/qj.589https://doi.org/10.1175/2010JTECHA1444.1https://doi.org/10.1175/2010JTECHA1444.1https://doi.org/10.1175/MWR-D-12-00362.1https://doi.org/10.1175/JTECH1910.1

an automated velocity dealiasing scheme for radar data ...22327...ational radar products for the...

Documents